Extreme genetic risk for type 1A diabetes
Theresa A. Aly†‡, Akane Ide†, Mohamed M. Jahromi†, Jennifer M. Barker†, Maria S. Fernando†, Sunanda R. Babu†,
Liping Yu†, Dongmei Miao†, Henry A. Erlich§, Pamela R. Fain†‡, Katherine J. Barriga†, Jill M. Norris¶,
Marian J. Rewers†¶, and George S. Eisenbarth†‡储
†Barbara Davis Center for Childhood Diabetes and ‡Human Medical Genetics Program, University Colorado Health Sciences Center, Aurora, CO 80045;
§Roche Molecular Systems, Alameda, CA 94501; and ¶Department of Preventive Medicine and Biometrics, University of Colorado at Denver and
Health Sciences Center, Denver, CO 80262
Communicated by David W. Talmage, University of Colorado Health Sciences Center, Denver, CO, July 31, 2006 (received for review May 15, 2006)
Type 1A diabetes (T1D) is an autoimmune disorder the risk of which
is increased by specific HLA DR兾DQ alleles [e.g., DRB1*03-
DQB1*0201 (DR3) or DRB1*04-DQB1*0302 (DR4)]. The genotype
associated with the highest risk for T1D is the DR3兾4-DQ8 (DQ8 is
DQA1*0301, DQB1*0302) heterozygous genotype. We determined
HLA-DR and -DQ genotypes at birth and analyzed DR3兾4-DQ8
siblings of patients with T1D for identical-by-descent HLA haplo-
type sharing (the number of haplotypes inherited in common
between siblings). The children were clinically followed with
prospective measurement of anti-islet autoimmunity and for pro-
gression to T1D. Risk for islet autoimmunity dramatically increased
in DR3兾4-DQ8 siblings who shared both HLA haplotypes with their
diabetic proband sibling (63% by age 7, and 85% by age 15)
compared with siblings who did not share both HLA haplotypes
with their diabetic proband sibling (20% by age 15, P < 0.01). 55%
sharing both HLA haplotypes developed diabetes by age 12 versus
5% sharing zero or one haplotype (P ⴝ 0.03). Despite sharing both
HLA haplotypes with their proband, siblings without the HLA
DR3兾4-DQ8 genotype had only a 25% risk for T1D by age 12. The
risk for T1D in the DR3兾4-DQ8 siblings sharing both HLA haplotypes
with their proband is remarkable for a complex genetic disorder
and provides evidence that T1D is inherited with HLA-DR兾DQ
alleles and additional MHC-linked genes both determining major
risk. A subset of siblings at extremely high risk for T1D can now be
identified at birth for trials to prevent islet autoimmunity.
haplotype 兩 human leukocyte antigen 兩 major histocompatibility complex
A large body of evidence indicates that type 1A diabetes
(T1D) is an autoimmune disorder with important genetic
determinants, and it has become one of the most intensively
studied complex genetic disorders (1–3). The major histocom-
patibility complex (MHC) is reported to account for ⬇40% of
the familial aggregation of T1D (4, 5). The HLA-DR and -DQ
genes (linked HLA genes in the class II region of the MHC) are
well established as being associated with risk for T1D. Although
studies have implicated loci other than the HLA-DR and -DQ
loci (i.e., HLA-DPB1, HLA-A, HLA-B, and various non-HLA
genes) with diabetes risk and earlier age of onset, no loci with
contribution to risk equivalent to that of the HLA-DR and -DQ
alleles have been identified (6–10).
The insulin, PTPN22, and CTLA4 genes are non-HLA dia-
betes-susceptibility loci with allelic odds ratios of 1.9, 1.7, and 1.2,
respectively (11–14). However, even in combination with HLA
alleles, none of these identified loci confer a risk ⬎25% in
prospective studies. Nevertheless, a remaining fundamental
question is whether there are genetic polymorphisms other than
the HLA-DR and -DQ alleles that confer major risk for T1D. If
such loci existed, they could be linked to the MHC and would
thus be ‘‘hidden’’ in most linkage studies by the dramatic
Downloaded at Indonesia: PNAS Sponsored on January 24, 2020
influence of the HLA-DR and -DQ alleles.
Haplotypes are defined by sets of closely linked genetic
variants making chromosomal segments. In a family, the
inheritance of these chromosomal segments can be readily
defined, and, in particular, the number of haplotypes identical
by descent (shared) between siblings (none, one, or two)
14074 –14079 兩 PNAS 兩 September 19, 2006 兩 vol. 103 兩 no. 38
determined. Well over a decade ago, Glenys Thomson and
others pioneered the concept that affected sibling pairs are
more likely to share both HLA haplotypes than unaffected
sibling pairs (15–19). It was shown that children sharing two
haplotypes with an affected sibling have a risk of 16%, versus
9% for those sharing one, and ⬍1% for those sharing none
(20). This excess sharing of haplotypes in affected sibling pairs
could ref lect the inf luence of alleles at the DRB1 and DQB1
loci and, possibly, other MHC loci (21, 22). In the nonobese
diabetic mouse model, class I alleles in addition to class II
alleles contribute to diabetes (23, 24).
The Diabetes Autoimmunity Study of the Young (DAISY)
enrolled its first newborn subjects in 1993. Since then, cord blood
from more than 33,000 newborns has been screened for
HLA-DR and -DQ alleles associated with diabetes risk and
protection (25). Those at increased risk, including 1,058 unaf-
fected young relatives of patients with T1D, were enrolled for
follow-up detection of anti-islet autoantibodies and diabetes.
DAISY and other studies (26–28) have found that risk for T1D
in early childhood is higher in siblings than in offspring of
individuals with T1D, despite all having the same high-risk
HLA-DR and -DQ alleles.
We hypothesized that siblings of a child with diabetes have a
higher diabetes risk compared with offspring of a parent with
diabetes, even though siblings and offspring both share approx-
imately half of their genome with their diabetic proband, because
siblings can share both HLA haplotypes identical to their
proband, whereas offspring inherit only one haplotype from
their single diabetic parent. Specifically, the sharing of multiple
genetic polymorphisms of DR, DQ genes and non-DR, DQ
genes linked to the MHC region on both copies of chromosome
6 could cause the increased sibling risk. We determined that we
could characterize the amount of risk contributed by non-DR,
DQ loci by analyzing the prospective risk associated with HLA
haplotype sharing in siblings that all had the same high-risk
DR3兾4-DQ8 (DQ8 is DQA1*0301, DQB1*0302) genotype. We
analyzed genotypes of sibling families (proband, sibling, and
parents) at the HLA class II region and, if needed, at additional
polymorphisms within the MHC to determine the number of
HLA haplotypes shared between siblings and their diabetic
probands.
Here, we report a remarkable high risk for DR3兾4-DQ8
children who share both HLA haplotypes identical by descent
Author contributions: T.A.A., A.I., M.M.J., K.J.B., J.M.N., M.J.R., and G.S.E. designed re-
search; T.A.A., A.I., M.M.J., M.S.F., S.R.B., L.Y., and D.M. performed research; T.A.A., A.I.,
M.M.J., J.M.B., M.S.F., S.R.B., P.R.F., and G.S.E. analyzed data; H.A.E. contributed new
reagents兾analytic tools; K.J.B. database preparation and management; and T.A.A. and
G.S.E. wrote the paper.
The authors declare no conflict of interest.
Freely available online through the PNAS open access option.
Abbreviations: T1D, type 1A diabetes; DR3, DRB1*03-DQB1*0201; DR4, DRB1*04-
DQB1*0302; DRB1*03-DQB1*0201兾DRB1*04-DQB1*0302 (DR3兾4-DQ8); DAISY, The Diabe-
tes Autoimmunity Study of the Young.
储To
whom correspondence should be addressed. E-mail: george.eisenbarth@uchsc.edu.
© 2006 by The National Academy of Sciences of the USA
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0606349103
 Fig. 1. Haplotype sharing analysis. Haplotype sharing is illustrated in three representative families in which the DR3兾4-DQ8 DAISY sibling shares both (family
A), one (family B), or no (family C) haplotypes with the diabetic proband.
with their affected sibling. These results are consistent with the
effect of a gene(s) that confers major risk for T1D that is linked
to, but distinct from, the HLA-DR and -DQ alleles.
Results
We studied a group of children with the highest risk HLA
genotype, DR3兾4-DQ8 heterozygotes. By life table analysis,
DR3兾4-DQ8 siblings in the study had a 41% (⫾8% SEM) overall
risk for developing islet autoantibodies by age 7, compared
with a 16% (⫾6%) risk in DR3兾4-DQ8 offspring (P ⫽ 0.01) and
a 5% (⫾1.6%) risk in DR3兾4-DQ8 children from the general
population.
Fig. 1 illustrates three representative DAISY families where a
DAISY sibling was prospectively monitored for islet autoimmu-
nity and development of diabetes. All three DAISY siblings have
the DR3兾4-DQ8 genotype. The DAISY sibling in family A
shares both haplotypes with the sibling proband, with the
proband defined as the first child in the family diagnosed with
T1D. The DAISY sibling in family B shares one haplotype with
the sibling proband, and the DAISY sibling in family C does not
share any haplotypes with the sibling proband. DAISY siblings
being followed in this study of 48 families always have the
DR3兾4-DQ8 alleles, but the diabetic proband sibling does not
always have the DR3, DQ2 or DR4, DQ8 alleles. Even if the
diabetic proband has DR3, DQ2 and DR4, DQ8 alleles, they may
be inherited from different parental haplotypes than those
inherited by the DR3兾4-DQ8 sibling.
The majority of the DR3兾4-DQ8 DAISY siblings shared both
HLA haplotypes with the proband (n ⫽ 29), 17 siblings shared
one haplotype, one sibling shared no haplotypes, and haplotype
sharing was either zero or one for 1 sibling (family was not
Downloaded at Indonesia: PNAS Sponsored on January 24, 2020
further informative with the markers analyzed). We performed
survival-curve analysis to compare progression to islet autoan-
tibody positivity (Fig. 1, A) and diabetes (Fig. 1, B) in DR3兾4-
DQ8 DAISY siblings, stratified by the number of haplotypes
they shared with the sibling proband (Fig. 2). The survival curves
in Fig. 2 A show that DR3兾4-DQ8 siblings sharing both HLA
Aly et al.
haplotypes with their diabetic proband sibling have a 63%
(⫾11% SEM) risk of developing persistent anti-islet autoanti-
bodies by age 7 and an 85% (⫾12%) risk by age 15, versus a 20%
(⫾11%) risk by age 15 for those not sharing both haplotypes. To
date, 55% (16 of 29) sharing both haplotypes versus 16% (3 of
19) sharing zero or one haplotype are autoantibody positive (P ⫽
0.008). DR3兾4-DQ8 siblings sharing both HLA haplotypes with
their diabetic proband sibling have a 55% (⫾13%) risk for
diabetes (Fig. 2B) by age 12, compared with a 7% (⫾7%) risk for
those not sharing both haplotypes (Fig. 2B). To date, 34% (10
of 29) versus 5% (1 of 19) have type 1 diabetes (P ⫽ 0.03). The
eight remaining nondiabetic DR3兾4-DQ8 DAISY siblings that
are persistently positive for anti-islet autoantibodies are at high
risk for diabetes given their high-risk HLA alleles and their
expression of multiple anti-islet autoantibodies, with two or
more anti-islet autoantibodies expressed in five of these siblings
(Table 1).
For DAISY siblings sharing both haplotypes with their pro-
band but lacking the DR3兾4-DQ8 genotype (n ⫽ 27), risk by
survival-curve analysis is only 23% for anti-islet autoantibodies
at age 15 and 25% for diabetes at age 12. These results indicate
that high-risk DR-DQ alleles (e.g., DR3-DQ2 and DR4-DQ8)
and HLA haplotype sharing with the proband sibling are both
essential for extremely high diabetes risk.
We summarize the disease status of each DAISY DR3兾4-DQ8
GENETICS
sibling and list the DRB1 and DQB1 alleles (including DRB1*04
subtypes), haplotypes shared with probands, HLA-DPB1, and
HLA-A alleles in Table 1. When there was only one haplotype
inherited in common by the siblings, the DR3-DQ2 haplotype
was shared in 41% of the siblings (n ⫽ 7) and the DR4-DQ8
haplotype was shared in 59% (n ⫽ 10) (Table 1, rightmost
column). The DRB1*04 subtypes did not differ significantly
between the groups stratified by haplotype sharing. Because
specific alleles at the HLA-DPB1 locus have an association with
diabetes risk (29, 30), we analyzed genotypes of the DR3兾4-DQ8
siblings at the DPB1 locus. The high-risk DPB1 alleles were not
significantly associated with diabetic autoimmunity in this group
PNAS 兩 September 19, 2006 兩 vol. 103 兩 no. 38 兩 14075
 Fig. 2. Haplotype sharing survival curves. Shown is progression to anti-islet autoimmunity (A) and type 1A diabetes (B) in DR3兾4-DQ8 siblings stratified by the
number of HLA haplotypes shared with their proband siblings. n ⫽ 48 for A and B; error bars for all panels represent the SEM.
(n ⫽ 47; Table 2). Similarly, the distribution of the HLA-A alleles
(HLA-A*0201 and HLA-A*24 alleles) did not significantly
differ between groups (6). Thus, MHC genes previously reported
as being associated with T1D do not explain the magnitude of
increased risk observed with HLA haplotype sharing.
Higher maternal age at delivery, male gender, and younger
proband sibling age at diagnosis are risk factors that were
associated with sibling recurrence risk for diabetes in a study of
⬎10,000 sibling pairs in the Finnish population (31). We per-
formed Cox proportional hazards regression analysis on the
DR3兾4-DQ8 sibling data to evaluate risk for islet autoimmunity
and diabetes, using HLA haplotype sharing, maternal age at
delivery, gender, and age of proband at diagnosis as potential
explanatory variables. Only HLA haplotype sharing was associ-
ated with risk in the DR3兾4-DQ8 siblings using the likelihood
ratio statistic (n ⫽ 48, P ⫽ 0.03 for islet autoimmunity, P ⫽ 0.04
for diabetes) and none of the other variables contributed sig-
nificantly. However, the results of this analysis of the DAISY
siblings do not refute the findings of the Finnish study, be-
cause the Finnish study (of ⬎10,000 siblings pairs) had more
power to detect factors associated with minor risk than our
more selective study of DAISY siblings with the DR3兾4-DQ8
genotype (n ⫽ 48).
The DR3兾4-DQ8 siblings were also genotyped at HLA-B (n ⫽
47) to identify alleles associated with extended haplotypes (e.g.,
HLA-B*08 and -B*15). The presence of alleles associated with
the HLA-DR3-B8-A1 extended haplotype in DR3兾4-DQ8 sib-
lings that shared both HLA haplotypes with their proband sibling
did not significantly add to their risk (n ⫽ 28, 31% with alleles
B8 and A1 affected versus 25% unaffected, P ⫽ 1). In a similar
way, the presence of the HLA B15 allele was not significantly
increased in affected siblings that shared both haplotypes (n ⫽
28, 13% affected versus 33% unaffected, P ⫽ 0.36).
We analyzed the HphI insulin gene polymorphism to assess its
effect on risk in the DR3兾4-DQ8 siblings that shared both
haplotypes. Of the DR3兾4-DQ8 siblings with the highest-risk
insulin genotype (AA) who shared both HLA haplotypes with
the proband, 47% (⫾14%) progressed to autoantibody positivity
Downloaded at Indonesia: PNAS Sponsored on January 24, 2020
by age 2.5 (Fig. 3A) and 56% (⫾17%) to diabetes by age 7.5 (Fig.
3B). By age 5, a significantly higher number of DR3兾4-DQ8
siblings with the high-risk insulin genotype who shared both
HLA haplotypes with the proband developed T1D versus sib-
lings with the lower-risk insulin genotypes, excluding nondiabetic
siblings younger than age 5 (5 of 11 versus 0 of 12, P ⫽ 0.01).
14076 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0606349103
Thus, the high-risk insulin genotype may be associated with a
lower age-of-onset of T1D in DR3兾4-DQ8 children sharing both
haplotypes with a diabetic sibling. We also analyzed risk asso-
ciated with the PTPN22 gene, but the high-risk allele, T, did not
significantly influence risk for anti-islet autoantibody develop-
ment in the DAISY siblings studied [CT or TT genotype in 16%
of affected (n ⫽ 19) and 21% of unaffected (n ⫽ 28) siblings].
We sought a different population of DR3兾4 siblings for initial
confirmation of the added importance of HLA haplotype shar-
ing to the DR3-DQ2兾DR4-DQ8 genotype. The Human Biolog-
ical Data Interchange database has families selected a priori for
multiple diabetic siblings; however, unlike the DAISY siblings,
the Human Biological Data Interchange siblings were not pro-
spectively followed to diabetes. The proband was defined as the
first sibling in the family to develop diabetes. We found that 83%
(35 of 42) of the DR3兾4-DQ8 siblings sharing both haplotypes
by descent with their diabetic proband were diabetic versus 58%
(18 of 31) not sharing both haplotypes with their diabetic
proband (n ⫽ 73, P ⫽ 0.03).
Discussion
We report the discovery of a genetic risk for type T1D three to
four times higher than previously reported (32). A feature of this
study is the simple combined analysis of the HLA-DR and -DQ
highest-risk alleles with analysis of HLA haplotype sharing. The
DAISY study has followed children prospectively from birth for
development of anti-islet autoantibodies and diabetes with HLA
typing of ⬎33,000 newborns to aid in risk stratification (26–28).
Among those HLA typed, only 29 siblings have the DR3兾4-DQ8
alleles and share both HLA haplotypes identical by descent with
their diabetic proband sibling. Nevertheless, ⬇20% of the chil-
dren currently diagnosed with T1D from the DAISY population
are in this tiny group of 29 children. By survival-curve analysis,
this group has a 63% risk of progressing to persistent anti-islet
autoimmunity by age 7 and a 55% risk of overt diabetes by age
12. The eventual risk for anti-islet autoimmunity for this sub-
group of siblings with further follow up is unknown but will
almost certainly exceed 80% (Fig. 2). This subgroup not only has
an enormous risk for T1D but also comprises a significant
percentage of the cases of early childhood anti-islet autoimmu-
nity identified in all participants of the large-scale DAISY study.
Our preliminary results indicate that an even more rapid onset
of islet autoimmunity may occur in DR3兾4-DQ8 siblings that
Aly et al.
 Table 1. Summary of Phenotype, Haplotype Sharing, and HLA Genotypes
Positive No. Haplotypes
ID T1D AutoAb DQ DRB1*04 DQ DR shared shared DPB1 HLA-A

483-0 Yes All 3 8 0401 2 3 2 Both 0101兾0401 03兾32

471-0 Yes All 3 8 0401 2 3 2 Both 0401兾0201 01兾02
667-0 Yes All 3 8 0402 2 3 2 Both 0201兾0401 24兾66
60-0 Yes All 3 8 0404 2 3 2 Both 0201兾0401 01兾02
539-0 Yes All 3 8 0402 2 3 2 Both 0401兾0501 02兾03
533-0 Yes IAA, IA-2 8 0401 2 3 2 Both 0101兾0301 02兾03
132-0 Yes IAA, GAD 8 0401 2 3 2 Both 0301兾1601 01兾02
330-0 Yes IAA, GAD 8 0401 2 3 2 Both 0301兾1601 01兾02
45–0 Yes IAA, IA-2 8 0401 2 3 2 Both 0201兾0401 01兾02
275-0 Yes IAA 8 0405 2 3 2 Both 0101兾0201 29兾01 or 36
343-0 Yes IAA 8 0405 2 3 1 DQ2 0301兾0401 02兾03
40-0 No All 3 8 0401 2 3 2 Both 0202兾0401 01兾03
399-0 No All 3 8 0401 2 3 2 Both 0301兾0401 02兾30
41-0 No GAD, IA-2 8 0401 2 3 2 Both 0401兾0401 02兾02
18-0 No GAD 8 0401 2 3 2 Both 0201兾0401 01兾01
133-0 No IA-2 8 0401 2 3 2 Both 0301兾0401 01兾02
656-0 No IAA 8 0404 2 3 2 Both 0401兾0201 02兾02
409-0 No All 3 8 0401 2 3 1 DQ2 0401兾0401 01兾02
732-0 No All 3 8 0401 2 3 1 DQ8 0201兾0201 02兾29
21-0 No 0 8 0401 2 3 2 Both 0101兾0401 02兾32
178-0 No 0 8 0401 2 3 2 Both 0301兾0401 03兾24
182-0 No 0 8 0401 2 3 2 Both 0401兾0401 01兾02
263-0 No 0 8 0401 2 3 2 Both 0101兾0401 01兾11
325-0 No 0 8 0401 2 3 2 Both 0101兾0402 02兾25
707-0 No 0 8 0401 2 3 2 Both 0101兾0201 02兾24
764-0 No 0 8 0401 2 3 2 Both 0201兾0401 01/ 03 or 11
811-0 No 0 8 0401 2 3 2 Both 0202兾0401 02兾30
815-0 No 0 8 0401 2 3 2 Both 0101兾0401 01兾11
810-0 No 0 8 0404 2 3 2 Both 0202兾0402 02兾30
858-0 No 0 8 0404 2 3 2 Both
958-0 No 0 8 0404 2 3 2 Both 0301兾0401 11兾23
1020-0 No 0 8 0404 2 3 2 Both 0101兾2001 03兾26
352-0 No 0 8 0401 2 3 1 DQ2 0101兾0101 02兾25
938-0 No 0 8 0401 2 3 1 DQ2 0201兾0401 02兾02
922-0 No 0 8 0401 2 3 1 DQ2 0301兾0402 01兾29
457-0 No 0 8 0401 2 3 1 DQ8 0401兾0901 02兾03
731-0 No 0 8 0401 2 3 1 DQ8 0201兾0201 02兾29
736-0 No 0 8 0401 2 3 1 DQ8 0201兾0401 03兾31
800-0 No 0 8 0401 2 3 1 DQ8 0401兾0401 01兾02
580-0 No 0 8 0401 2 3 1 DQ8 0401兾0401 02兾25
255-0 No 0 8 0404 2 3 1 DQ8 0301兾0401 01兾24
925-0 No 0 8 0405 2 3 1 DQ8 0401兾0401 02兾02
397-0 No 0 8 0404 2 3 0 or 1 Unknown 0301兾0401 01兾01
281-0 No 0 8 0403 2 3 1 DQ8 0401兾0401 02兾03
759-0 No 0 8 0407 2 3 1 DQ2 0301兾0402 01兾33
767-0 No 0 8 0407 2 3 1 DQ2 0401兾0401 02兾68
321-0 No 0 8 0408 2 3 1 DQ8 1301兾1501 01兾03
538-0 No 0 8 0401 2 3 0 Neither 0101兾0201 02兾03

Abbreviations: T1D, type 1A diabetes; Positive AutoAb, positive islet autoantibodies; # Shared, number of haplotypes shared between siblings.
GENETICS

have the high-risk genotype for the insulin gene in addition to
sharing both HLA haplotypes with their diabetic proband.
This study also illuminates the most probable explanation for
the ‘‘relative paradox’’ of siblings having higher diabetes risk
compared with offspring and the general population, despite all
having matched high-risk DR-DQ genotypes. One possible
Downloaded at Indonesia: PNAS Sponsored on January 24, 2020
other proposed explanation has been that siblings have increased
risk due to the polygenic inheritance of T1D, with multiple loci
outside of the MHC contributing to risk. However, neither of
these two initial explanations accounts for the dramatic differ-
ence in risk for DAISY DR3兾4-DQ8 siblings sharing both HLA
haplotypes versus those not sharing both haplotypes with their

affected probands and confirmed by our independent analysis of

explanation for the increased risk of siblings has been that
‘‘family based’’ shared environmental factors are a major con-
tributor to disease. However, even though environmental factors
may be crucial, if they are common (for instance, all individuals
exposed to a triggering virus or dietary factor), they would not
significantly contribute to familial aggregation of disease. An-
DR3兾4-DQ8 Human Biological Data Interchange siblings. Our
results provide strong evidence to support the hypothesis that
siblings, unlike offspring, can inherit both HLA haplotypes
identical by descent with their sibling proband, implicating
MHC-linked non-DR, DQ genes plus high-risk DR and DQ
alleles.

Aly et al. PNAS 兩 September 19, 2006 兩 vol. 103 兩 no. 38 兩 14077
 Table 2. DPB1 allele frequencies curves in this article are based on prospective follow-up of both
Ab⫺ Ab⫹ Diabetic siblings and general population newborns from birth.
Haplotypes, Haplotypes, subjects, subjects, patients, First-degree relatives (n ⫽ 1,058) and a subset of the general
DPB1 n % % % % population children (n ⫽ 1,411) were followed prospectively for
the development of anti-islet autoantibodies. Of children en-
0101 12 12.8 14.0 7.9 13.6 rolled for follow-up, 49 full siblings and 51 offspring of T1D
0201* 16 17.0 12.3 23.7 22.7 relatives and 398 general population children carried the highest-
0202* 3 3.2 3.5 2.6 0.0 risk HLA-DR3兾4-DQ8 genotype. The DAISY study also en-
0301* 12 12.8 10.5 15.8 18.2 rolled 319 non-HLA-DR3兾4-DQ8 siblings of T1D relatives.
0401 40 42.6 45.6 42.1 31.8 Insulin, GAD65, and IA-2 autoantibody levels were measured at
0402** 4 4.3 7.0 0.0 0.0 follow-up visits at 9, 15, and 24 months of age and annually
1601 2 2.1 0.0 5.3 9.1 thereafter in children with normal autoantibody levels (27). We
Other 5 5.3 7.0 2.6 4.5 obtained and analyzed parental, proband, and sibling DNA from
Total 94.0 100.0 100.0 100.0 100.0 48 families of the 49 DR3兾4-DQ8 siblings of patients with T1D
Reported as higher (*) and lower (**) risk. DPB1 allele frequencies in 47 to directly determine the number of HLA haplotypes shared
DR3兾4-DQ8 siblings of diabetic probands. Ab⫺, islet autoantibody negative; identical by descent between siblings.
Ab⫹, islet autoantibody positive. To further test the influence of haplotype sharing for DR3兾
4-DQ8 siblings, we analyzed available data from families of the
Human Biological Data Interchange collection. From 273 fam-
The higher risk in children that share both HLA haplotypes ilies, we identified 73 DR 3兾4-DQ8 siblings of diabetic proband
with their diabetic siblings also suggests that the susceptibility children. The number of DR3兾4-DQ8 children with diabetes
allele or alleles may be inherited in a recessive manner (33). sharing both haplotypes with their diabetic proband siblings (n ⫽
Regardless of the model of inheritance, it is likely that HLA- 42) was compared with those sharing zero or one haplotypes
DR3-DQ2 and -DR4-DQ8 haplotypes shared by descent be- (n ⫽ 31).
tween affected siblings contain a higher proportion of high-risk
alleles than HLA-DR3-DQ2 and -DR4-DQ8 haplotypes not Genotyping. HLA-DRB1, HLA-DQB1, HLA-B, HLA-A, and
shared by descent. We believe that the search for additional HLA-DP genotyping of the DAISY siblings was performed with
polymorphisms contributing to young-onset anti-islet autoim- immobilized sequence-specific oligonucleotide genotyping sim-
munity should concentrate on detailed analysis of genes within ilar to a previously described methodology (36). Family members
or linked to the MHC. other than the DAISY siblings were genotyped at the HLA-A
It may be much easier to prevent anti-islet autoimmunity than locus only if one or more parents were homozygous at the
to prevent progression to diabetes once anti-islet autoimmunity HLA-DR and HLA-DQ loci to determine which HLA haplo-
is initiated, as indicated by studies of animal models (34). Infants types were inherited and shared.
and children who have the highest genetic risk for diabetes as Siblings of probands from these families were also genotyped at
defined by this study are a major group that may now be the insulin and PTPN22 loci by using previously published methods
considered for initial clinical trials to prevent childhood diabetes (12, 37). The insulin polymorphism was detected by sequencing at
before the development of anti-islet autoantibodies. the HphI locus. The PTPN22 polymorphism was detected by XcmI
restriction enzyme digestion of amplified DNA.
Methods
Population. DAISY is a prospective study in which blood samples Statistical Analysis. The statistical difference between the survival
from ⬎33,000 young first-degree relatives of patients with T1D curves for progression to anti-islet autoantibody positivity or
and general population newborns were obtained with informed diabetes in DAISY children was evaluated with the log-rank test
parental consent and institutional review board oversight at the with Prism software (Prism Software Corporation, Irvine, CA).
University of Colorado Health Sciences Center (27, 35). Survival The Fisher two-sided exact test was used to compare the number
Downloaded at Indonesia: PNAS Sponsored on January 24, 2020

Fig. 3. Haplotype sharing and insulin genotype survival curves. Shown is progression to anti-islet autoantibody positivity (A) and type 1 diabetes (B) in
DR3兾4-DQ8 siblings sharing both HLA haplotypes with their sibling probands stratified by genotypes of the insulin gene at the HphI polymorphism. The survival
curves represent the high-risk genotypes (AA; n ⫽ 13) versus the lower-risk genotypes (AT, TT; n ⫽ 14). Error bars represent the SEM.

14078 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0606349103 Aly et al.

 of individuals who developed persistent anti-islet autoantibodies
and diabetes in the DR3兾4-DQ8 siblings who shared both
haplotypes versus those who did not share both haplotypes. The
Fisher two-sided exact test was also used to evaluate data for
uneven distribution of DRB1*04 subtypes, HLA-DP alleles,
HLA-A alleles, and HLA-B alleles, and for risk associated with
the insulin and PTPN22 loci in the DAISY siblings being
studied. The alpha level for significance for the Fisher exact test
was set at 0.05.
Cox proportional hazards regression analysis was performed
on the DAISY sibling data with HLA haplotype sharing, ma-
ternal age at delivery, gender, and age at diagnosis of the
1. Onengut-Gumuscu S, Concannon P (2002) Immunol Rev 190:182–194.
2. Anjos S, Polychronakos C (2004) Mol Genet Metab 81:187–195.
3. Pugliese A (2004) Endocrinol Metab Clin North Am 33:1–16, vii.
4. Noble JA, Valdes AM, Cook M, Klitz W, Thomson G, Erlich HA (1996) Am J
Hum Genet 59:1134–1148.
5. Lambert AP, Gillespie KM, Thomson G, Cordell HJ, Todd JA, Gale EA,
Bingley PJ (2004) J Clin Endocrinol Metab 89:4037–4043.
6. Nakanishi K, Kobayashi T, Murase T, Nakatsuji T, Inoko H, Tsuji K, Kosaka
K (1993) Diabetes 42:1086–1098.
7. Dawkins RL, Christiansen FT, Kay PH, Garlepp M, McCluskey J, Holling-
sworth PN, Zilko PJ (1983) Immunol Rev 70:1–22.
8. Valdes AM, Erlich HA, Noble JA (2005) Hum Immunol 66:301–313.
9. Robinson WP, Barbosa J, Rich SS, Thomson G (1993) Genet Epidemiol
10:273–288.
10. Johansson S, Lie BA, Todd JA, Pociot F, Nerup J, Cambon-Thomsen A,
Kockum I, Akselsen HE, Thorsby E, Undlien DE (2003) Genes Immun
4:46–53.
11. Concannon P, Erlich HA, Julier C, Morahan G, Nerup J, Pociot F, Todd JA,
Rich SS (2005) Diabetes 54:2995–3001.
12. Bottini N, Muscumeci L, Alonso A, Rahmouni S, Nika K, Rostamkhani M,
MacMurray J, Meloni G, Lucarelli P, Pellechia M, et al. (2004) Nature Genet
36:337–338.
13. Smyth D, Cooper JD, Collins JE, Heward JM, Franklyn JA, Howson JM, Vella
A, Nutland S, Rance HE, Maier L, et al. (2004) Diabetes 53:3020–3023.
14. Ueda H, Howson JM, Esposito L, Heward J, Snook H, Chamberlain G,
Rainbow DB, Hunter KM, Smith AN, Di Genova G, et al. (2003) Nature
423:506–511.
15. Thomson G (1983) Tissue Antigens 21:81–104.
16. Thomson G, Robinson WP, Kuhner MK, Joe S, MacDonald MJ, Gottschall JL,
Barbosa J, Rich SS, Bertrams J, Baur MP, et al. (1988) Am J Hum Genet
43:799–816.
17. Valdes AM, Thomson G, Erlich HA, Noble JA (1999) Diabetes 48:1658–1661.
18. Cudworth AG, Woodrow JC (1975) Br Med J 3:133–135.
19. Deschamps I, Lestradet H, Busson M, Hors J (1986) Diabetes Res 3:391–396.
Downloaded at Indonesia: PNAS Sponsored on January 24, 2020
Aly et al.
proband as predictor variables. The likelihood ratio statistic was
used to test the significance of the variables with an alpha level
of significance set at 0.05.
We thank Brooke Hensley and Derrick Houser for laboratory assistance.
This work was supported by National Institutes of Health Diabetes
Autoimmunity Study of the Young (DAISY) Grants DK32083 and
DK32493, Autoimmunity Prevention Center Grant AI50964, Diabetes
Endocrine Research Center Grant P30DK57516, Clinical Research
Centers Grants M01 RR00069 and M01 RR00051, Immune Toler-
ance Network Grant AI 15416, the American Diabetes Association, the
Juvenile Diabetes Research Foundation, and the Children’s Diabetes
Foundation.
20. Tarn AC, Thomas JM, Dean BM, Ingram D, Schwarz G, Bottazzo GF, Gale
EA (1988) Lancet 1:845–850.
21. Blomhoff A, Lie BA, Myhre AG, Kemp EH, Weetman AP, Akselsen HE,
Huseby ES, Undlien DE (2004) J Clin Endocrinol Metab 89:3474–3476.
22. Hanifi MP, de Knijf P, Roep BO, van der AB, Naipal A, Gorus F, Schuit F,
Giphart MJ (1998) Diabetes 47:263–269.
23. Inoue K, Ikegami H, Fujisawa T, Noso S, Nojima K, Babaya N, Itoi-Babaya M,
Makimo S, Ogihara T (2004) Diabetologia 47:739–747.
24. Hattori M, Yamato E, Itoh N, Senpuku H, Fujisawa T, Yoshino M, Fukuda M,
Matsumoto E, Toyonaga T, Nakagawa I, et al. (1999) J Immunol 163:1721–
1724.
25. Barker JM, Goehrig SH, Barriga K, Hoffman M, Slover R, Eisenbarth GS,
Norris JM, Klingensmith GJ, Rewers M (2004) Diabetes Care 27:1399–1404.
26. Achenbach P, Koczwara K, Knopff A, Naserke H, Ziegler AG, Bonifacio E
(2004) J Clin Invest 114:589–597.
27. Barker JM, Barriga KJ, Yu L, Miao D, Erlich HA, Norris JM, Eisenbarth GS,
Rewers M (2004) J Clin Endocrinol Metab 89:3896–3902.
28. Hermann R, Bartsocas CS, Soltesz G, Vazeou A, Paschou P, Bozas E,
Malamitsi-Puchner A, Simell O, Knip M, Ilonen J (2004) Diabetes Metab Res
Rev 20:322–329.
29. Noble JA, Valdes AM, Thomson G, Erlich HA (2000) Diabetes 49:121–125.
30. Nishimaki K, Kawamura T, Inada H, Yagawa K, Nose Y, Nabeya N, Isshiki G,
Tatsumi N, Niihira S (2000) Diabetes Res Clin Pract 47:49–55.
31. Harjutsalo V, Podar T, Tuomilehto J (2005) Diabetes 54:563–569.
32. Tarn AC, Thomas JM, Dean BM, Ingram D, Schwarz G, Bottazzo GF, Gale
EA (1988) Lancet 1:845–850.
33. Larsen CE, Alper CA (2004) Curr Opin Immunol 16:660–667.
34. Atkinson MA, Leiter EH (1999) Nat Med 5:601–604.
35. Rewers M, Bugawan TL, Norris JM, Blair A, Beaty B, Hoffman M, McDuffie
RS, Jr, Hamman RF, Klingensmith G, Eisenbarth GS, et al. (1996) Diabetologia
39:807–812.
36. Bugawan TL, Erlich HA (1991) Immunogenetics 33:163–170.
37. Pugliese A, Miceli D (2002) Diabetes Metab Res Rev 18:13–25.
GENETICS
PNAS 兩 September 19, 2006 兩 vol. 103 兩 no. 38 兩 14079